How Important Is DeFragging?

[Mike Chapman]

It’s a dirty little secret that most shared-storage manufacturers introduce fragmentation to keep performance linear over the capacity of the drive. That’s why it’s usually not a good idea to defragment a media volume – your performance might die in the middle of the sequence, depending on where your now-contiguous files end up. If they’re at the outside of the disk platter, performance goes into the tank. I don’t know if fragmentation is built into the editing apps themselves; all I do know is that Avid especially used to issue warnings about defragging your media drives.

Mike Chapman

[Eddie Lotter]

If you are not experiencing performance issues then it’s not very important for you.

Defragmnenting a drive rearranges files on the drive so that each file is written contiguously as opposed to being written in fragments all over the disk.

Fragmentation makes the computer work much harder in reading a file back and in video editing that can cause stuttering during playback.

It’s not a bad idea to put the defrag on a weekly schedule and let it run overnight.

Cheers
Eddie

[Tim Kolb]

[Mike Chapman] “all I do know is that Avid especially used to issue warnings about defragging your media drives.”

Interesting. In those days we had Media 100s and Defragmented…and in Apple’s case…Optimized (not only creating contiguous files, but moving them to create contiguous free space) regularly.

Avid’s specific file system may have had some issue with defrag as back in the old days, Macs could only see 4GB partitions in the OS (I assume that the same sector/cluster limitations that eventually led Windows to move to NTFS were the barrier, but I’m guessing) and Avid may have had a proprietary way of formatting media drives to get around that which would have been messed up by an OS defragmentation scheme…I know RISC computers have never had the same limitations.

As far as speed goes, technically, the fastest data access would be on the outside of the platter. Physics says the relative area passing under the head with a constant spindle speed will be significantly more at the outer edge than near the center spindle.

However,

AFAIK, most harddrives actually write data less densely on the outer tracks to equalize speed with the tracks closer to the spindle and to make search/access faster as the sectors line up as the head scans in and out making it unnecessary to read each track/sector/cluster to tell where it is. It really ends up being a circular projection of a grid with longitude and latitude if you will…

The sectors are written in longer stretches as the tracks get closer to the outside so the data is spread out a bit more…the sector boundaries then line up and if one could see it, the sector boundaries would make the disk platter look as if it was sliced like a pie, from center to outer edge, each track segment in the ‘slice’ being a sector with each sector holding the same amount of clusters whether the most inner track with the smallest relative area, or the outer track with exponentially more area.

So…I’m not sure how performance would decrease based on strictly position on the platter unless the sectors were laid in a linear manner instead of angular, and even then the outer tracks would be much faster I/O than the inner tracks. Though I would suspect that with linear sectors, search time would end up being absurd…the disk would act like a fast LTO tape with linear search to at least a large extent.

…so in a nutshell, the assertion that a disk is faster in the center perplexes me.

TimK,
Director, Consultant
Kolb Productions,